As reported by The Hacker News, security researchers have published a fully working exploit — dubbed AnyPwn — for a pre-authentication remote code execution vulnerability in AnyDesk's Linux client that yields root privileges before any user approves the incoming connection. The flaw was patched in June with version 8.0.3, but AnyDesk's changelog described it only as a crash fix. No CVE has been assigned. No security advisory was published. This is exactly the kind of disclosure failure that forces defenders to play catch-up months after a fix ships.
Vulnerability Profile
| Field | Detail |
|---|---|
| Vendor / Product | AnyDesk for Linux |
| Affected Versions | 8.0.2 and likely earlier (8.0.1 unconfirmed) |
| Patched Version | 8.0.3 (June 2026); latest is 8.1.0 |
| CVE Identifier | None assigned as of October 9, 2026 |
| CVSS / Severity | Not rated (Shield53 assessment: Critical — pre-auth, network-reachable, root-level RCE) |
| Exploit Available | Yes — public PoC on GitHub (released October 8, 2026) |
| Active Exploitation | Not yet confirmed in the wild, but public PoC dramatically increases risk |
| Attack Vector | Direct TCP on port 7070; relay path demonstrated via Frida but full chain unconfirmed |
Why This Matters
The technical root cause is a textbook 32-bit integer overflow in AnyDesk's mode-5 stream packet handler. The attacker declares a payload length of 0xFFFFFFF0; the handler adds a 16-byte header (0x10), wrapping the 32-bit result to zero. The allocator reserves a tiny buffer, and the subsequent copy overwrites adjacent heap memory. The published PoC achieves reliable enough heap manipulation to redirect execution — all before the connection-approval dialog appears on the target screen.
What elevates this beyond a single-product bug is the disclosure posture. AnyDesk shipped a patch, labeled it a crash fix, assigned no CVE, issued no advisory, and reportedly removed the vulnerable 8.0.2 build from its download page after the PoC video surfaced. This pattern — silent patching followed by build scrubbing — leaves enterprise defenders blind. Asset management systems that rely on CVE feeds show nothing. Vulnerability scanners that match on CVE IDs report clean. Patch-management pipelines that prioritize based on severity ratings deprioritize or skip the update entirely. The result is a four-month window where Linux endpoints running AnyDesk 8.0.2 or earlier sit exposed to a pre-auth root compromise with a public exploit available.
The absence of a CVE is not an absence of risk. It is an absence of visibility.
Who Is Most Exposed
- Linux endpoints with AnyDesk directly exposed on TCP 7070 — especially jump hosts, support technician workstations, and unattended servers in IT/OT environments.
- MSPs and helpdesk operations that rely on AnyDesk for remote support and may not have enforced the June update.
- Organizations with permissive firewall rules allowing inbound 7070 from broad IP ranges rather than specific management subnets.
- Cloud-hosted Linux instances with AnyDesk installed for convenience access where port 7070 is reachable from the internet.
The Relay Question
AnyDesk stated the vulnerability is limited to direct connections and does not affect relay paths. However, the researchers demonstrated that the same vulnerable code path is reachable through relay servers using Frida instrumentation, though they did not complete a full exploit chain over relays. If relay exploitation proves feasible, the exposure surface expands dramatically — any AnyDesk Linux client reachable through the vendor's relay infrastructure could be a target, regardless of local firewall posture on port 7070. Defenders should treat the relay path as potentially exploitable until AnyDesk provides a formal technical justification for why it is not.
Shield53 Recommendations
Immediate Actions
anydesk and any bundled deployment mechanisms.Strategic Actions
- Establish a vendor disclosure watchlist. Track changelog language for critical remote-access tools (AnyDesk, TeamViewer, RustDesk, ScreenConnect) for terms like "crash fix," "stability improvement," or "memory handling" that may mask security patches. Cross-reference with researcher disclosures and GitHub PoC repositories.
- Request a formal CVE from AnyDesk through your support channel or CISA's CVE program. The absence of a CVE for a pre-auth root RCE with a public exploit is a significant gap that hampers industry-wide risk management.
- Reassess remote-access tool architecture. Consider whether direct TCP exposure is ever necessary, or whether all remote support can operate through relay or VPN-tunneled paths with additional authentication layers.
This incident underscores a recurring problem in the remote-access tooling space: vendors that treat security fixes as optional disclosures. When a pre-authentication root-level RCE ships as a "crash bug" with no CVE, the entire defensive ecosystem — from asset inventories to patch prioritization to threat modeling — breaks down. The fix is available. The exploit is public. The question is whether your environment has already applied the one and is ready for the other.
